Latest news with #Git
Business Wire
4 days ago
- Business
- Business Wire
Reddit Accelerates Innovation and Slashes Build Times with Buildkite
SAN FRANCISCO--(BUSINESS WIRE)-- Buildkite Pty Ltd., a leader in scalable software delivery solutions, today announced that Reddit, one of the world's largest online communities with more than over 400M weekly active users, has successfully migrated its mobile continuous integration and delivery (CI/CD) pipelines to Buildkite, dramatically improving build performance, reliability, and developer autonomy across its mobile engineering organization. By adopting Buildkite's high-performance hosted agents and dynamic pipeline primitives, Reddit has reduced build queue times to just seconds and accelerated build speeds by 30% for both Android and iOS workflows, while maintaining cost efficiency. By adopting Buildkite's high-performance hosted agents and dynamic pipeline primitives, Reddit has reduced build queue times to just seconds and accelerated build speeds by 30% for both Android and iOS workflows, while maintaining cost efficiency. Share 'With Buildkite, we finally have the reliability and performance at scale that our mobile teams were missing,' said Ken Struys, Reddit's Director of Developer Experience. 'We've seen a significant improvement to the overall developer experience, and both Android and iOS CI are more than 30% faster on Buildkite.' Facing growing demands from millions of global users and a rapidly expanding codebase, Reddit's mobile teams had outgrown their previous CI/CD provider. Lack of ability to control the build environment meant they couldn't use custom Docker images (as they can with Buildkite). The team also struggled with configuration complexity, with 6,000 lines of YAML split across multiple files, in addition to unreliable builds due to environment drift and dependency failures. Reddit's migration to Buildkite, affecting over 200 mobile engineers, was completed ahead of schedule and involved a full transition of both Android and iOS mono repositories. The company leveraged Buildkite's composable pipeline primitives to create custom workflows that seamlessly integrated with Bazel and BuildBuddy, allowing both iOS and Android apps to be built efficiently on Linux runners with remote execution. Dynamic pipelines enabled Reddit to build powerful CI systems with less cruft and less code repetition. Comprehensive benchmarking revealed that Reddit's builds now run up to 30% faster, with job queue times consistently dropping to around five seconds. Buildkite's Git caching and container caching features were 'absolute game changers,' reducing Git checkout times from several minutes to 30-40 seconds. Intelligent build cancellation and log customization further improved developer experience and operational efficiency. The migration was conducted by a remarkably lean core team. 'We built most of the Android pipelines and the iOS pipelines with two people,' said Struys. 'You don't need massive engineering resources to use Buildkite, but the scale is tremendous. We've got 170 plus engineers on it now, and most of our builds are running like 12 to 15 jobs in each field now.' It was also executed with exceptional care through a staged approach, including having shadow builds running in parallel to production systems to ensure a zero-risk transition. Buildkite's team provided responsive support and delivered custom feature requests, including a GitHub App for GitHub Enterprise Server, throughout the process. "Reddit's successful migration exemplifies the transformative power of modern, composable CI/CD architecture," said Dan Ring, Vice President of Product at Buildkite, who worked closely with the company on the migration. "When engineering organizations outgrow traditional CI/CD limitations, they need a platform that can deliver both immediate performance gains and long-term scalability. Reddit's experience demonstrates what's possible when you combine Buildkite's high-performance hosted agents with our flexible, composable primitives. This isn't just about faster builds, it's about unleashing engineering teams to innovate at the speed their users demand." For more information, see Reddit's blog post here. About Buildkite Pty Ltd: Based in San Francisco and Melbourne, Buildkite is a fast-growing software delivery provider that offers the industry's first and only Scale-Out Delivery Platform. Buildkite's Scale-Out Delivery platform is the only solution that provides the flexibility and scale required by the world's most demanding companies for delivering software across a broad range of use cases, including AI/ML workloads and mobile application development. Global innovation leaders including Airbnb, Block, Canva, Cruise, Culture Amp, Elastic, Lyft, PagerDuty, Pinterest, PlanetScale, Rippling, Shopify, Slack, Tinder, Twilio, Uber, and Wayfair have standardized on Buildkite for software delivery. For more information, please visit
Forbes
29-07-2025
- Business
- Forbes
5 Ways AI Agents Will Aid Human Engineers
Naveen Edapurath Vijayan is a Sr Manager of Data Engineering at AWS, specializing in data analytics and large-scale data systems. Not long ago, AI in engineering meant automated scripts or basic predictive analytics—useful, but hardly transformational. Today, we're entering a new era: AI agents that don't just support engineers but collaborate with them, learn from them and even mentor them. These agents are emerging as tireless, adaptable teammates. Fine-tuned on company-specific codebases, infrastructure and tribal knowledge, they mirror real engineering workflows and proactively enhance them. They won't replace engineers but will make them significantly more effective. For leaders, the opportunity is clear: the faster you learn how to integrate these agents into your teams, the faster you'll see compounding gains in speed, quality and knowledge sharing. Let's explore five ways AI agents are reshaping engineering teams and what leaders can do to embrace this evolution. Code Generation That Learns Your Stack In many organizations, onboarding new engineers takes months. They need to learn the architecture, read through layers of legacy code and understand your team's design patterns. Now imagine an AI agent that's already absorbed your entire Git history. One that can generate working components in your tech stack, offer suggestions consistent with your code style and even surface context from past pull requests. I recently trialed such an agent on a backend service rewrite. The AI, fine-tuned on internal repositories, proposed boilerplate logic, caught inconsistencies during review and explained why certain deprecated modules were used in legacy branches. In effect, it did what an experienced mid-level developer would—only faster. The takeaway? Leaders should consider building internal fine-tuning pipelines early. Public large language models (LLMs) are great generalists, but the real value emerges when you let them learn your company's DNA. Start with smaller, low-risk services before scaling to critical production systems. Automated Test Writing And Coverage One of the least glamorous yet most critical parts of engineering is testing. Many teams underinvest in tests because they're tedious and hard to maintain. AI agents now bridge that gap. Trained on both source code and historical test suites, they can write meaningful unit and integration tests, not just syntactically correct ones, but tests that mirror the edge cases your team has historically encountered. In one case, I have seen an AI assistant uncover a dormant API regression by auto-generating a test case similar to one that had failed two years prior. It didn't just generate tests. It learned from patterns of breakage. The business value here is straightforward: reduced bugs in production, faster CI/CD cycles and less developer fatigue. For leaders, the shift is cultural as much as technical: treat testing as an AI-augmented activity and prioritize feeding historical test failures and bug reports into these agents. Even a 10%-20% improvement in test coverage compounds quickly in production reliability. SQL And Data Pipeline Assistants Data engineers often spend too much time writing boilerplate SQL, debugging pipelines or interpreting undocumented joins across dozens of tables. Now imagine an AI assistant trained on your schema, glossary and business metrics, one that can suggest joins, flag inefficiencies and translate business questions into query templates. In some data-heavy migrations, such an assistant cut onboarding time for new analysts by half, helping junior staff navigate complex ETL logic without scheduling handoffs or Slacking someone in another time zone. If you're leading a data organization, think of your company's schema as a knowledge base. The more structured and documented it is, the better these AI copilots will perform. Leaders should invest in cleaning and governing data definitions, not just for compliance but because it directly improves how well these tools can accelerate work. Infrastructure As Code And DevOps Agents Modern infrastructure is programmable, which means it's also teachable. AI agents are now being trained on Terraform scripts, deployment logs and incident reports. They can monitor for anomalies, recommend rollbacks or even auto-generate provisioning code based on architectural intent. The key insight here: AI doesn't have to be reactive. With historical telemetry and clear boundaries, it can become your first responder and, in some cases, your reliability engineer. Leaders should start by feeding agents well-structured incident postmortems and defining strict guardrails for automated changes. Begin with staging environments and use AI to augment, not replace, human approvals until reliability is proven. Knowledge Capture And Onboarding Bots One of the biggest productivity killers in engineering is lost context. Decisions in meetings, Slack conversations or tribal knowledge rarely make it into documentation. AI agents now capture, structure and resurface this institutional memory. An onboarding assistant trained on Confluence, Asana and Slack can answer, 'Why did we migrate off Kafka last year?' by surfacing the relevant thread, decision memo and postmortem within seconds. The value is clear: smoother onboarding, fewer repeated mistakes and less reliance on senior engineers. Leaders should map 'tribal' knowledge and make it accessible to agents. The Road Ahead We're entering an era where engineers will increasingly design systems with AI, not just for AI. These agents won't replace the creativity, judgment or ethical oversight of human developers, but they will make every developer faster, more informed and more resilient. However, leaders must also plan for the risks: Hallucinations and overconfidence: Even fine-tuned agents can generate plausible but wrong outputs—human review is still essential. Data leakage and IP risks: Strict access controls are critical when letting agents learn from proprietary code. Cultural pushback: Engineers may distrust AI suggestions unless you invest in training and change management. Skill atrophy: Over-reliance on AI for routine tasks can erode core engineering skills if not balanced with intentional learning. According to a 2024 McKinsey analysis, generative AI is driving disruption across the software value chain, reshaping developer productivity, tooling choices and the economics of software delivery. Similarly, Gartner highlights that by 2026, generative AI will be embedded into 80% of development tools—fundamentally transforming how software is built and maintained. For forward-thinking leaders, the message is clear: Don't ask if AI will replace engineers. Ask how you can reimagine your engineering teams with AI as a powerful ally. Forbes Technology Council is an invitation-only community for world-class CIOs, CTOs and technology executives. Do I qualify?
Geeky Gadgets
24-07-2025
- Geeky Gadgets
Pro Claude Code Workflow for Seamless Software Development
Have you ever wondered what separates a chaotic coding process from a seamless, productive workflow? In the fast-paced world of software development, the tools you choose—and how you use them—can make all the difference. Enter Claude Code, a powerful AI-driven assistant designed to elevate your development experience. But here's the catch: unlocking its full potential requires more than just surface-level familiarity. To truly harness its capabilities, you need a structured approach that aligns with both your technical goals and collaborative needs. From streamlining testing to optimizing memory and managing context, Claude offers a suite of features that can transform the way you code. In this guide, Yifan explains 6 Claude Code workflows, breaking down its essential components and advanced functionalities. You'll discover how to automate repetitive tasks, maintain clarity in complex projects, and foster seamless collaboration across teams. Whether you're looking to integrate version control with Git, customize permissions for efficiency, or master advanced commands for intricate workflows, this guide will equip you with actionable insights to enhance productivity and code quality. As you read on, consider this: what could your development process look like if every tool and feature worked in perfect harmony? Claude Code Workflow Guide Streamlining Testing for Reliable Development Testing is a cornerstone of dependable software development, and Claude Code offers robust support for integrating testing into your workflow. By embedding automated tests during the prototyping phase, you can ensure functionality and reliability from the outset. Claude can assist in generating or executing tests, making it easier to verify code behavior. Incorporating frameworks such as unit testing or integration testing tools further enhances precision and efficiency. To maintain consistent code quality, automate pre-commit checks like linting, compiling, and running tests. These checks help identify issues early in the development lifecycle, reducing the risk of errors in production. By streamlining testing, you create a stable foundation for iterative development and long-term project success. Optimizing Memory Management for Clarity Efficient memory management is essential for maintaining clarity and consistency in your projects. Claude Code uses a ` file to store project context, rules, and guidelines. Regularly updating this file using the `/init` command ensures that Claude remains aligned with the current state of your project. However, it is crucial to avoid overloading the file with unnecessary details, as this can overwhelm the system and degrade performance. Focus on including only essential guidelines and context to keep the workflow streamlined. By managing memory effectively, you can ensure that Claude delivers accurate and relevant responses, even in complex projects. Complete Claude Code Workflow Watch this video on YouTube. Unlock more potential in Claude Code Workflow by reading previous articles we have written. Effective Context Handling for Accurate Outputs Managing context is critical to making sure that Claude provides accurate and relevant outputs. After completing a task, use commands like `/clear` or `/compact` to remove outdated or irrelevant information. This practice prevents confusion and keeps the system focused on current objectives. If you need to revisit previous discussions, the `claude d-res` command allows you to navigate back to specific points in the conversation. While Claude can assist with context management, it is advisable to handle version control manually to maintain consistency across iterations. This approach ensures that your codebase remains organized and free from discrepancies, even as projects evolve. Customizing Permissions for Streamlined Workflows Claude Code's permission settings offer flexibility to tailor workflows according to your needs. Modes such as read-only, auto-accept edits, and plan mode provide varying levels of control, allowing you to adapt the system to different scenarios. For example, the 'dangerously skip permissions' mode can accelerate workflows but should only be used in controlled environments to minimize risks. Adjusting default permission settings can further enhance efficiency by reducing unnecessary prompts and interruptions. By customizing permissions, you can create a smoother, more streamlined development process. Integrating Version Control with Git Integrating Git with Claude Code simplifies version control and ensures that your codebase remains organized. Automating pre-commit checks—such as compiling, linting, and running tests—helps maintain a stable repository. Additionally, Claude can generate commit messages based on the context of your changes, saving time and effort during the development process. This integration supports iterative development by providing a clear history of changes and facilitating collaboration among team members. By using version control effectively, you can maintain a clean and reliable codebase. Enhancing Collaboration Across Teams Claude Code supports asynchronous collaboration, making it an ideal tool for distributed teams. Integration with platforms like GitHub enables efficient pull request reviews and issue-based changes. Customizing prompts for concise and relevant code reviews ensures that feedback is actionable and easy to implement. Automated workflows, such as continuous integration and testing, further enhance collaboration by reducing manual effort and making sure consistency. These features allow teams to work effectively across time zones, improving overall productivity and project outcomes. Using Utility Features for Productivity Claude Code includes a variety of utility features designed to boost productivity and simplify workflows. Key functionalities include: Adding memory rules with the `#` command to maintain project guidelines. Executing bash commands using `!` for quick terminal operations. Adjusting problem-solving intensity with keywords like 'think' or 'ultra think.' Undoing terminal inputs with `Ctrl + -` for error correction. Analyzing images by pasting them directly into the interface for visual insights. For projects involving multiple repositories, referencing additional directories can streamline workflows. Non-interactive mode is particularly useful for generating quick outputs without verbose logs, making it ideal for rapid prototyping and debugging. Using Advanced Commands for Complex Workflows Advanced commands in Claude Code simplify complex workflows and enhance efficiency. For example: Use `claude d-res` to revisit specific conversation points and maintain continuity. Resume previous discussions seamlessly with `claude d-continue` for ongoing tasks. Streamline debugging and troubleshooting with non-interactive modes to focus on outputs. These commands, when combined with structured workflows and effective context management, enable you to fully use Claude's capabilities. By mastering these advanced features, you can tackle complex projects with confidence and precision. Maximizing Claude Code's Potential A structured and strategic approach to Claude Code can significantly enhance your development workflow. By focusing on key areas such as testing, memory management, context handling, permissions, version control, and collaboration, you can achieve higher productivity and maintain exceptional code quality. Using utility features and advanced commands further refines your processes, allowing you to work efficiently in any development environment. Media Credit: Yifan Filed Under: AI, Guides Latest Geeky Gadgets Deals Disclosure: Some of our articles include affiliate links. If you buy something through one of these links, Geeky Gadgets may earn an affiliate commission. Learn about our Disclosure Policy.
Techday NZ
18-07-2025
- Business
- Techday NZ
AWS brings vibe coding to the Enterprise with spec-driven Kiro IDE tool
AWS has introduced Kiro, an "agentic IDE" designed to bridge the gap between the excitement of prompt-based prototyping and the practical demands of production software. According to Kiro product lead Nikhil Swaminathan, the tool aims to bring structure, rigour and automation to the modern, AI-powered coding process. Swaminathan describes the appeal of recent AI tools, saying, "Prompt, prompt, prompt, and you have a working application. It's fun and feels like magic. But getting it to production requires more." He outlines the typical stumbling blocks: "What assumptions did the model make when building it? What edge-cases did it cover? How did it handle errors? Requirements are fuzzy and you can't tell if the application meets them." Kiro is designed to solve these problems by introducing what its creators call "spec-driven development." As the team puts it, "Kiro is great at 'vibe coding' but goes way beyond that - Kiro's strength is getting those prototypes into production systems with features such as specs and hooks." Swaminathan explains how it works: "Start with a prompt: 'Add a review system for products.' Kiro translates this into a set of user stories with EARS-style acceptance criteria." He says Kiro then generates artefacts including "a data-flow diagram, TypeScript interfaces, a database schema, and API definitions." The system's approach includes automatically specifying essential features for each user story. Swaminathan writes, "Kiro automatically includes requirements like mobile responsiveness, accessibility, loading states, and tests in the spec." Tasks are then "sequenced correctly and connected to requirements." Importantly, the specs remain in sync as the code evolves. Swaminathan notes, "Developers can author code and ask Kiro to update specs or manually update specs to refresh tasks. This solves the common problem where developers stop updating original artifacts during implementation." To automate repetitive work, Kiro introduces "agent hooks." These are "event-driven automations" that "trigger based on events like file saves or deletions." As Swaminathan puts it, "When you save a React component, hooks update the test file. When you modify API endpoints, hooks refresh README files. When you're ready to commit, security hooks scan for leaked credentials." He describes the benefit: "It's like having an experienced developer catching things you miss or completing boilerplate tasks." These hooks are also collaborative by design. Swaminathan explains, "Once this hook is committed to Git, it enforces the coding standard across my entire team - whenever anyone adds a new component, the agent automatically validates it against the guidelines." Kiro is built on top of Code OSS, meaning it is "compatible with existing VS Code settings and Open VSX plugins." It supports "Model Context Protocol (MCP)," agentic chat, and multiple context providers, including "files, URLs and document uploads." Looking ahead, Swaminathan and AWS VP of Developer Experience & Agents Deepak Singh set out an ambitious vision for Kiro. They write, "We want to tackle the root causes of pain in software development - clarity of design, alignment with requirements, technical debt, code reviews, and knowledge sharing." Kiro is available in a preview release for Mac, Windows and Linux, supporting most programming languages. Swaminathan invites developers to experience its approach: "We invite you to try Kiro and share feedback. We're just getting started, and your input will help shape the future of agentic development." By combining the "magic" of AI-powered coding with structured specs and event-driven automation, Kiro is positioning itself as a tool for developers seeking to move quickly without sacrificing discipline or reliability.
Techday NZ
18-07-2025
- Business
- Techday NZ
AWS brings vibe coding to the Enterpise with spec-driven Kiro IDE tool
AWS has introduced Kiro, an "agentic IDE" designed to bridge the gap between the excitement of prompt-based prototyping and the practical demands of production software. According to Kiro product lead Nikhil Swaminathan, the tool aims to bring structure, rigour and automation to the modern, AI-powered coding process. Swaminathan describes the appeal of recent AI tools, saying, "Prompt, prompt, prompt, and you have a working application. It's fun and feels like magic. But getting it to production requires more." He outlines the typical stumbling blocks: "What assumptions did the model make when building it? What edge-cases did it cover? How did it handle errors? Requirements are fuzzy and you can't tell if the application meets them." Kiro is designed to solve these problems by introducing what its creators call "spec-driven development." As the team puts it, "Kiro is great at 'vibe coding' but goes way beyond that - Kiro's strength is getting those prototypes into production systems with features such as specs and hooks." Swaminathan explains how it works: "Start with a prompt: 'Add a review system for products.' Kiro translates this into a set of user stories with EARS-style acceptance criteria." He says Kiro then generates artefacts including "a data-flow diagram, TypeScript interfaces, a database schema, and API definitions." The system's approach includes automatically specifying essential features for each user story. Swaminathan writes, "Kiro automatically includes requirements like mobile responsiveness, accessibility, loading states, and tests in the spec." Tasks are then "sequenced correctly and connected to requirements." Importantly, the specs remain in sync as the code evolves. Swaminathan notes, "Developers can author code and ask Kiro to update specs or manually update specs to refresh tasks. This solves the common problem where developers stop updating original artifacts during implementation." To automate repetitive work, Kiro introduces "agent hooks." These are "event-driven automations" that "trigger based on events like file saves or deletions." As Swaminathan puts it, "When you save a React component, hooks update the test file. When you modify API endpoints, hooks refresh README files. When you're ready to commit, security hooks scan for leaked credentials." He describes the benefit: "It's like having an experienced developer catching things you miss or completing boilerplate tasks." These hooks are also collaborative by design. Swaminathan explains, "Once this hook is committed to Git, it enforces the coding standard across my entire team - whenever anyone adds a new component, the agent automatically validates it against the guidelines." Kiro is built on top of Code OSS, meaning it is "compatible with existing VS Code settings and Open VSX plugins." It supports "Model Context Protocol (MCP)," agentic chat, and multiple context providers, including "files, URLs and document uploads." Looking ahead, Swaminathan and AWS VP of Developer Experience & Agents Deepak Singh set out an ambitious vision for Kiro. They write, "We want to tackle the root causes of pain in software development - clarity of design, alignment with requirements, technical debt, code reviews, and knowledge sharing." Kiro is available in a preview release for Mac, Windows and Linux, supporting most programming languages. Swaminathan invites developers to experience its approach: "We invite you to try Kiro and share feedback. We're just getting started, and your input will help shape the future of agentic development." By combining the "magic" of AI-powered coding with structured specs and event-driven automation, Kiro is positioning itself as a tool for developers seeking to move quickly without sacrificing discipline or reliability.
